Re: [Intel-gfx] [PATCH CI 2/4] drm/i915/display/psr: Use plane damage clips to calculate damaged area

[Daniel Vetter <daniel@ffwll.ch>]

On Wed, Jul 21, 2021 at 9:56 AM Daniel Vetter <daniel@ffwll.ch> wrote:
> On Wed, Jul 21, 2021 at 9:50 AM Daniel Vetter <daniel@ffwll.ch> wrote:
> > On Tue, Jul 20, 2021 at 6:55 PM Souza, Jose <jose.souza@intel.com> wrote:
> > > On Tue, 2021-07-20 at 17:31 +0200, Daniel Vetter wrote:
> > > > On Tue, Jul 20, 2021 at 5:16 PM Daniel Vetter <daniel@ffwll.ch> wrote:
> > > > > On Tue, Jul 20, 2021 at 5:09 PM Daniel Vetter <daniel@ffwll.ch> wrote:
> > > > > > On Mon, Jan 4, 2021 at 9:56 PM José Roberto de Souza
> > > > > > <jose.souza@intel.com> wrote:
> > > > > > >
> > > > > > > Now using plane damage clips property to calcualte the damaged area.
> > > > > > > Selective fetch only supports one region to be fetched so software
> > > > > > > needs to calculate a bounding box around all damage clips.
> > > > > > >
> > > > > > > Now that we are not complete fetching each plane, there is another
> > > > > > > loop needed as all the plane areas that intersect with the pipe
> > > > > > > damaged area needs to be fetched from memory so the complete blending
> > > > > > > of all planes can happen.
> > > > > > >
> > > > > > > v2:
> > > > > > > - do not shifting new_plane_state->uapi.dst only src is in 16.16 format
> > > > > > >
> > > > > > > v4:
> > > > > > > - setting plane selective fetch area using the whole pipe damage area
> > > > > > > - mark the whole plane area damaged if plane visibility or alpha
> > > > > > > changed
> > > > > > >
> > > > > > > v5:
> > > > > > > - taking in consideration src.y1 in the damage coordinates
> > > > > > > - adding to the pipe damaged area planes that were visible but are
> > > > > > > invisible in the new state
> > > > > > >
> > > > > > > v6:
> > > > > > > - consider old state plane coordinates when visibility changes or it
> > > > > > > moved to calculate damaged area
> > > > > > > - remove from damaged area the portion not in src clip
> > > > > > >
> > > > > > > v7:
> > > > > > > - intersec every damage clip with src to minimize damaged area
> > > > > > >
> > > > > > > v8:
> > > > > > > - adjust pipe_damaged area to 4 lines grouping
> > > > > > > - adjust calculation now that is understood that uapi.src is the
> > > > > > > framebuffer coordinates that plane will start to fetch from
> > > > > > >
> > > > > > > v9:
> > > > > > > - Only add plane dst or src to damaged_area if visible
> > > > > > > - Early skip plane damage calculation if it was not visible in old and
> > > > > > > new state
> > > > > > >
> > > > > > > Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
> > > > > > > Cc: Gwan-gyeong Mun <gwan-gyeong.mun@intel.com>
> > > > > > > Reviewed-by: Gwan-gyeong Mun <gwan-gyeong.mun@intel.com>
> > > > > > > Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
> > > > > >
> > > > > > Why is this not using drm_atomic_helper_damage_merged? I just stumbled
> > > > > > over this, and this is one of the only two drivers that directly digs
> > > > > > around in the damage area, and seems to reinvent a bunch of the stuff
> > > > > > here.
> > >
> > > We can use drm_atomic_helper_damage_merged() but it would only save us one for loop.
> >
> > Yes please. The trouble with rolling our own copies for everything is
> > that it does add up.
> >
> > > > > Also, did we merge the igts for this stuff? They unfortunately never
> > > > > landed, when vmwgfx team did all this work, but for i915 we really
> > > > > shouldn't even land new support without tests.
> > > >
> > > > Lo and behold, we merge the uapi enabling way earlier than this patch here:
> > > >
> > > > commit 093a3a30000926b8bda9eef773e4ed5079053350
> > > > Author: José Roberto de Souza <jose.souza@intel.com>
> > > > Date:   Thu Jun 25 18:01:47 2020 -0700
> > > >
> > > >    drm/i915: Add plane damage clips property
> > > >
> > > > And the igts are nowhere near to be seen, at least the stuff from
> > > > vmwgfx didn't land. Please file a JIRA internally and ping me on that
> > > > so this gets sorted out asap.
> > >
> > > Here the IGT: https://gitlab.freedesktop.org/drm/igt-gpu-tools/-/blob/master/tests/kms_psr2_sf.c
> >
> > There was some igts that were cross-driver, not intel specific. The
> > thing here is supposed to be a cross-vendor interface, so would be
> > really good if you resurrect those from vmwgfx folks and land them
> > too:
> >
> > https://patchwork.freedesktop.org/series/51087/
>
> Also from a very quick look at the kms_psr2_sf tests there's really
> not a whole lot specific about our psr2 implementation in there. The
> test should work on any plane with a FB_DAMAGE_CLIPS rect available,
> so there's no reason to make this specific to intel, much less to
> psr2. I think we need to:
> - make this a generic kms test, it should work
> - merge with the testcases from the folks who merged FB_DAMAGE_CLIPS
> originally to make sure all drivers follow the same contract.
>
> KMS properties are generic, intel-specific tests for when there's
> nothing intel-specific isn't good. Also adding Pankaj.

Looks like Pankaj no longer working at intel.

Another one is this here:

    igt_debug_manual_check("all", expected);

That's like the worst name because it means we can only enable this
manual checkpoint when "all" is set, which also enables all others. It
seems copypasta from from other kms_psr tests. Please:
- fix this and replace the string with something useful
- patch the igt functions igt_debug_manual_check() and
igt_debug_wait_for_keypress() to assert if "all" is supplied, and
update docs accordingly

I'm also not super convinced the testcase design makes sense, it
creates a new subtest for each block position, instead of having a few
functional subtests and then moving the blocks around. FB_DAMAGE_CLIPS
is about updates, and you don't test updates if you split each update
into a separate subtest (which CI splits up in its runs). That just
wastes machine time.

If this is copypasta from other psr tests, then those need to be fixed up too.

Also since we're deep in igt discussions, adding igt list too.

Cheers, Daniel

> > >
> > > >
> > > > Thanks, Daniel
> > > >
> > > > > > > ---
> > > > > > >  drivers/gpu/drm/i915/display/intel_psr.c | 113 ++++++++++++++++++++---
> > > > > > >  1 file changed, 99 insertions(+), 14 deletions(-)
> > > > > > >
> > > > > > > diff --git a/drivers/gpu/drm/i915/display/intel_psr.c b/drivers/gpu/drm/i915/display/intel_psr.c
> > > > > > > index d9a395c486d3..f5b9519b3756 100644
> > > > > > > --- a/drivers/gpu/drm/i915/display/intel_psr.c
> > > > > > > +++ b/drivers/gpu/drm/i915/display/intel_psr.c
> > > > > > > @@ -1242,9 +1242,11 @@ static void psr2_man_trk_ctl_calc(struct intel_crtc_state *crtc_state,
> > > > > > >         if (clip->y1 == -1)
> > > > > > >                 goto exit;
> > > > > > >
> > > > > > > +       drm_WARN_ON(crtc_state->uapi.crtc->dev, clip->y1 % 4 || clip->y2 % 4);
> > > > > > > +
> > > > > > >         val |= PSR2_MAN_TRK_CTL_SF_PARTIAL_FRAME_UPDATE;
> > > > > > >         val |= PSR2_MAN_TRK_CTL_SU_REGION_START_ADDR(clip->y1 / 4 + 1);
> > > > > > > -       val |= PSR2_MAN_TRK_CTL_SU_REGION_END_ADDR(DIV_ROUND_UP(clip->y2, 4) + 1);
> > > > > > > +       val |= PSR2_MAN_TRK_CTL_SU_REGION_END_ADDR(clip->y2 / 4 + 1);
> > > > > > >  exit:
> > > > > > >         crtc_state->psr2_man_track_ctl = val;
> > > > > > >  }
> > > > > > > @@ -1269,8 +1271,8 @@ int intel_psr2_sel_fetch_update(struct intel_atomic_state *state,
> > > > > > >                                 struct intel_crtc *crtc)
> > > > > > >  {
> > > > > > >         struct intel_crtc_state *crtc_state = intel_atomic_get_new_crtc_state(state, crtc);
> > > > > > > +       struct drm_rect pipe_clip = { .x1 = 0, .y1 = -1, .x2 = INT_MAX, .y2 = -1 };
> > > > > > >         struct intel_plane_state *new_plane_state, *old_plane_state;
> > > > > > > -       struct drm_rect pipe_clip = { .y1 = -1 };
> > > > > > >         struct intel_plane *plane;
> > > > > > >         bool full_update = false;
> > > > > > >         int i, ret;
> > > > > > > @@ -1282,13 +1284,25 @@ int intel_psr2_sel_fetch_update(struct intel_atomic_state *state,
> > > > > > >         if (ret)
> > > > > > >                 return ret;
> > > > > > >
> > > > > > > +       /*
> > > > > > > +        * Calculate minimal selective fetch area of each plane and calculate
> > > > > > > +        * the pipe damaged area.
> > > > > > > +        * In the next loop the plane selective fetch area will actually be set
> > > > > > > +        * using whole pipe damaged area.
> > > > > > > +        */
> > > > > > >         for_each_oldnew_intel_plane_in_state(state, plane, old_plane_state,
> > > > > > >                                              new_plane_state, i) {
> > > > > > > -               struct drm_rect *sel_fetch_area, temp;
> > > > > > > +               struct drm_rect src, damaged_area = { .y1 = -1 };
> > > > > > > +               struct drm_mode_rect *damaged_clips;
> > > > > > > +               u32 num_clips, j;
> > > > > > >
> > > > > > >                 if (new_plane_state->uapi.crtc != crtc_state->uapi.crtc)
> > > > > > >                         continue;
> > > > > > >
> > > > > > > +               if (!new_plane_state->uapi.visible &&
> > > > > > > +                   !old_plane_state->uapi.visible)
> > > > > > > +                       continue;
> > > > > > > +
> > > > > > >                 /*
> > > > > > >                  * TODO: Not clear how to handle planes with negative position,
> > > > > > >                  * also planes are not updated if they have a negative X
> > > > > > > @@ -1300,23 +1314,94 @@ int intel_psr2_sel_fetch_update(struct intel_atomic_state *state,
> > > > > > >                         break;
> > > > > > >                 }
> > > > > > >
> > > > > > > -               if (!new_plane_state->uapi.visible)
> > > > > > > -                       continue;
> > > > > > > +               num_clips = drm_plane_get_damage_clips_count(&new_plane_state->uapi);
> > > > > > >
> > > > > > >                 /*
> > > > > > > -                * For now doing a selective fetch in the whole plane area,
> > > > > > > -                * optimizations will come in the future.
> > > > > > > +                * If visibility or plane moved, mark the whole plane area as
> > > > > > > +                * damaged as it needs to be complete redraw in the new and old
> > > > > > > +                * position.
> > > > > > >                  */
> > > > > > > -               sel_fetch_area = &new_plane_state->psr2_sel_fetch_area;
> > > > > > > -               sel_fetch_area->y1 = new_plane_state->uapi.src.y1 >> 16;
> > > > > > > -               sel_fetch_area->y2 = new_plane_state->uapi.src.y2 >> 16;
> > > > > > > +               if (new_plane_state->uapi.visible != old_plane_state->uapi.visible ||
> > > > > > > +                   !drm_rect_equals(&new_plane_state->uapi.dst,
> > > > > > > +                                    &old_plane_state->uapi.dst)) {
> > > > > > > +                       if (old_plane_state->uapi.visible) {
> > > > > > > +                               damaged_area.y1 = old_plane_state->uapi.dst.y1;
> > > > > > > +                               damaged_area.y2 = old_plane_state->uapi.dst.y2;
> > > > > > > +                               clip_area_update(&pipe_clip, &damaged_area);
> > > > > > > +                       }
> > > > > > > +
> > > > > > > +                       if (new_plane_state->uapi.visible) {
> > > > > > > +                               damaged_area.y1 = new_plane_state->uapi.dst.y1;
> > > > > > > +                               damaged_area.y2 = new_plane_state->uapi.dst.y2;
> > > > > > > +                               clip_area_update(&pipe_clip, &damaged_area);
> > > > > > > +                       }
> > > > > > > +                       continue;
> > > > > > > +               } else if (new_plane_state->uapi.alpha != old_plane_state->uapi.alpha ||
> > > > > > > +                          (!num_clips &&
> > > > > > > +                           new_plane_state->uapi.fb != old_plane_state->uapi.fb)) {
> > > > > > > +                       /*
> > > > > > > +                        * If the plane don't have damaged areas but the
> > > > > > > +                        * framebuffer changed or alpha changed, mark the whole
> > > > > > > +                        * plane area as damaged.
> > > > > > > +                        */
> > > > > > > +                       damaged_area.y1 = new_plane_state->uapi.dst.y1;
> > > > > > > +                       damaged_area.y2 = new_plane_state->uapi.dst.y2;
> > > > > > > +                       clip_area_update(&pipe_clip, &damaged_area);
> > > > > > > +                       continue;
> > > > > > > +               }
> > > > > > > +
> > > > > > > +               drm_rect_fp_to_int(&src, &new_plane_state->uapi.src);
> > > > > > > +               damaged_clips = drm_plane_get_damage_clips(&new_plane_state->uapi);
> > > > > > > +
> > > > > > > +               for (j = 0; j < num_clips; j++) {
> > > > > > > +                       struct drm_rect clip;
> > > > > > > +
> > > > > > > +                       clip.x1 = damaged_clips[j].x1;
> > > > > > > +                       clip.y1 = damaged_clips[j].y1;
> > > > > > > +                       clip.x2 = damaged_clips[j].x2;
> > > > > > > +                       clip.y2 = damaged_clips[j].y2;
> > > > > > > +                       if (drm_rect_intersect(&clip, &src))
> > > > > > > +                               clip_area_update(&damaged_area, &clip);
> > > > > > > +               }
> > > > > > >
> > > > > > > -               temp = *sel_fetch_area;
> > > > > > > -               temp.y1 += new_plane_state->uapi.dst.y1;
> > > > > > > -               temp.y2 += new_plane_state->uapi.dst.y2;
> > > > > > > -               clip_area_update(&pipe_clip, &temp);
> > > > > > > +               if (damaged_area.y1 == -1)
> > > > > > > +                       continue;
> > > > > > > +
> > > > > > > +               damaged_area.y1 += new_plane_state->uapi.dst.y1 - src.y1;
> > > > > > > +               damaged_area.y2 += new_plane_state->uapi.dst.y1 - src.y1;
> > > > > > > +               clip_area_update(&pipe_clip, &damaged_area);
> > > > > > > +       }
> > > > > > > +
> > > > > > > +       if (full_update)
> > > > > > > +               goto skip_sel_fetch_set_loop;
> > > > > > > +
> > > > > > > +       /* It must be aligned to 4 lines */
> > > > > > > +       pipe_clip.y1 -= pipe_clip.y1 % 4;
> > > > > > > +       if (pipe_clip.y2 % 4)
> > > > > > > +               pipe_clip.y2 = ((pipe_clip.y2 / 4) + 1) * 4;
> > > > > > > +
> > > > > > > +       /*
> > > > > > > +        * Now that we have the pipe damaged area check if it intersect with
> > > > > > > +        * every plane, if it does set the plane selective fetch area.
> > > > > > > +        */
> > > > > > > +       for_each_oldnew_intel_plane_in_state(state, plane, old_plane_state,
> > > > > > > +                                            new_plane_state, i) {
> > > > > > > +               struct drm_rect *sel_fetch_area, inter;
> > > > > > > +
> > > > > > > +               if (new_plane_state->uapi.crtc != crtc_state->uapi.crtc ||
> > > > > > > +                   !new_plane_state->uapi.visible)
> > > > > > > +                       continue;
> > > > > > > +
> > > > > > > +               inter = pipe_clip;
> > > > > > > +               if (!drm_rect_intersect(&inter, &new_plane_state->uapi.dst))
> > > > > > > +                       continue;
> > > > > > > +
> > > > > > > +               sel_fetch_area = &new_plane_state->psr2_sel_fetch_area;
> > > > > > > +               sel_fetch_area->y1 = inter.y1 - new_plane_state->uapi.dst.y1;
> > > > > > > +               sel_fetch_area->y2 = inter.y2 - new_plane_state->uapi.dst.y1;
> > > > > > >         }
> > > > > > >
> > > > > > > +skip_sel_fetch_set_loop:
> > > > > > >         psr2_man_trk_ctl_calc(crtc_state, &pipe_clip, full_update);
> > > > > > >         return 0;
> > > > > > >  }
> > > > > > > --
> > > > > > > 2.30.0
> > > > > > >
> > > > > > > _______________________________________________
> > > > > > > Intel-gfx mailing list
> > > > > > > Intel-gfx@lists.freedesktop.org
> > > > > > > https://lists.freedesktop.org/mailman/listinfo/intel-gfx
> > > > > >
> > > > > >
> > > > > >
> > > > > > --
> > > > > > Daniel Vetter
> > > > > > Software Engineer, Intel Corporation
> > > > > > http://blog.ffwll.ch
> > > > >
> > > > >
> > > > >
> > > > > --
> > > > > Daniel Vetter
> > > > > Software Engineer, Intel Corporation
> > > > > http://blog.ffwll.ch
> > > >
> > > >
> > > >
> > >
> >
> >
> > --
> > Daniel Vetter
> > Software Engineer, Intel Corporation
> > http://blog.ffwll.ch
>
>
>
> --
> Daniel Vetter
> Software Engineer, Intel Corporation
> http://blog.ffwll.ch



-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx